正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Occlusion will be a big challenge for robots operating in dense, obstacle rich environments common in manufacturing. Using low-res cameras mounted on limbs you can train a motion policy to outperform high-res head mounted or externally mounted cameras for obstacle avoidance.

Simon Kalouche

10,525 subscribers

41,343 次观看 • 1 年前 •via X (Twitter)

教育科学技术

Anya Rossi• Live Now

Private livecam show

11 条评论

Simon Kalouche 的头像

Simon Kalouche1 年前

Full paper here:

Akool Inc 的头像

Akool Inc1 年前

AI has made it easy for anyone to create realistic avatar videos without expensive gear. AKOOL and Synthesia are top choices for AI-generated "talking photo" videos. Which one is best for your team?

involuntarily incelibate 的头像

involuntarily incelibate1 年前

Thanks. Now I want low res eyeballs in my fingertips

Simon Kalouche 的头像

Simon Kalouche1 年前

Here ya go

Medora AI, Inc. 的头像

Medora AI, Inc.1 年前

occlusion is not a camera problem. Though this is a fun thought, I wouldn't build this.

Jon Miller Schwartz 的头像

Jon Miller Schwartz1 年前

For someone who seems to be anti humanoid, you sure post about humanoids a bunch? 😄

Medora AI, Inc. 的头像

Medora AI, Inc.1 年前

Also occupation is a harder problem when things are moving very quickly. Especially in crowded areas where feedback may be chaotic.

Alex Thatcher 的头像

Alex Thatcher1 年前

Interesting.

Peyman Servati 的头像

Peyman Servati1 年前

Use @TexavieInc MarsWear to make robots aware of environment and tactile information like hunan 🚀

Dayder 的头像

Dayder1 年前

Amazing. While the humanoid form-factor is, and will be for a while, more convenient in a world build by humans and for human workers, why not remove some of its limitations? Like this better targeted vision, more freedom of movement/rotation, and many more.

Douglas P. Mercher 的头像

Douglas P. Mercher1 年前

Point clouds are generated more efficiently with radar

相关视频

🚨 BIG! Indian Army DEPLOYS Robotic MULE (Multi Utility Legged Equipment) along the LOC under the Anti-Infiltration grid. — Operates across tough terrain — Equipped with cameras for surveillance — Can carry equipment — Capable of operating with a mounted weapon

🚨 BIG! Indian Army DEPLOYS Robotic MULE (Multi Utility Legged Equipment) along the LOC under the Anti-Infiltration grid. — Operates across tough terrain — Equipped with cameras for surveillance — Can carry equipment — Capable of operating with a mounted weapon

Megh Updates 🚨™

74,438 次观看 • 6 个月前

We are excited to share our #CORL2024 paper on learning quadrotor obstacle avoidance from the visual stream of a single #eventcamera! Trained entirely in simulation! We demonstrate obstacle avoidance both in the dark and in a forest up to 5m/s. PDF: Video: Project page: Event cameras are sensors that output per-pixel-level intensity changes at microsecond latency resolution; they feature nearly zero motion blur and high dynamic range but produce a very large volume of events under significant ego-motion and further lack a high-fidelity continuous-time sensor model in simulation, making direct #sim2real transfer not possible. By leveraging depth prediction as a pretext task, we pre-train a reactive obstacle avoidance policy with “approximated” simulated events and then fine-tune the perception component with limited events-and-depth real-world data. This technique bridges the sim2real gap for #eventcameras! As at the current state, there is no continuous-time sensor model for event cameras, we hope that this work can finally spur future research leveraging simulation for training event-vision-based policies to create faster, agile robots! Kudos to Anish Bhattacharya, @marcocannic, Vijay Kumar Nikolai Matni UZH Science University of Zurich UZH Space Hub UZH IfI European Research Council (ERC) GRASP Laboratory Penn Engineering

We are excited to share our #CORL2024 paper on learning quadrotor obstacle avoidance from the visual stream of a single #eventcamera! Trained entirely in simulation! We demonstrate obstacle avoidance both in the dark and in a forest up to 5m/s. PDF: Video: Project page: Event cameras are sensors that output per-pixel-level intensity changes at microsecond latency resolution; they feature nearly zero motion blur and high dynamic range but produce a very large volume of events under significant ego-motion and further lack a high-fidelity continuous-time sensor model in simulation, making direct #sim2real transfer not possible. By leveraging depth prediction as a pretext task, we pre-train a reactive obstacle avoidance policy with “approximated” simulated events and then fine-tune the perception component with limited events-and-depth real-world data. This technique bridges the sim2real gap for #eventcameras! As at the current state, there is no continuous-time sensor model for event cameras, we hope that this work can finally spur future research leveraging simulation for training event-vision-based policies to create faster, agile robots! Kudos to Anish Bhattacharya, @marcocannic, Vijay Kumar Nikolai Matni UZH Science University of Zurich UZH Space Hub UZH IfI European Research Council (ERC) GRASP Laboratory Penn Engineering

Davide Scaramuzza

17,219 次观看 • 1 年前

*Competition 🌟 Pick your favourite mounted print from this video, and you could win it! 🎉😀 Comment below OR quote tweet, to be in with a chance to win! 😊 (I have 100s of signed and mounted 10x8 prints for sale here; 😊🐦❤️

*Competition 🌟 Pick your favourite mounted print from this video, and you could win it! 🎉😀 Comment below OR quote tweet, to be in with a chance to win! 😊 (I have 100s of signed and mounted 10x8 prints for sale here; 😊🐦❤️

Carl Bovis

50,908 次观看 • 1 年前

Vikram-1’s spectacular liftoff captured in slow motion 🚀 The instant Vikram-1 cleared the First Launch Pad at Sriharikota — over in a blink at full speed, here frame by frame. Ignition. Plume ballooning off the deck. The first few feet of a rocket bound for orbit. 🎥 High-speed cameras mounted at the pad #Vikram1 #MissionAagamanInAction

Vikram-1’s spectacular liftoff captured in slow motion 🚀 The instant Vikram-1 cleared the First Launch Pad at Sriharikota — over in a blink at full speed, here frame by frame. Ignition. Plume ballooning off the deck. The first few feet of a rocket bound for orbit. 🎥 High-speed cameras mounted at the pad #Vikram1 #MissionAagamanInAction

Skyroot Aerospace

62,570 次观看 • 19 天前

A Chinese robotics company unveiled a highly advanced, blue-eyed desktop companion robot designed to bring physical, empathetic AI directly to the workspace. Unlike traditional static smart speakers, this device utilizes eye-mounted cameras and multiple degrees of freedom to enable fluid neck motion, responsive head positioning, and real-time eye tracking for natural, human-like engagement.

A Chinese robotics company unveiled a highly advanced, blue-eyed desktop companion robot designed to bring physical, empathetic AI directly to the workspace. Unlike traditional static smart speakers, this device utilizes eye-mounted cameras and multiple degrees of freedom to enable fluid neck motion, responsive head positioning, and real-time eye tracking for natural, human-like engagement.

Wes Roth

33,414 次观看 • 3 个月前

🇮🇳 AI DATA RACE: Workers in Indian factories started to wear cameras on their heads, to film their motions so robots can be trained on these videos. Big robot companies will train their humanoid robots, on movement data from Indian sweatshops … Wild 1/

🇮🇳 AI DATA RACE: Workers in Indian factories started to wear cameras on their heads, to film their motions so robots can be trained on these videos. Big robot companies will train their humanoid robots, on movement data from Indian sweatshops … Wild 1/

Lord Bebo

532,714 次观看 • 4 个月前

It’s moving day for the MNPD Mounted Patrol Unit! The horses will now be housed in a barn at Peeler Park in Madison. As you can see, they’re already enjoying all the space and fresh grass in their new pasture.

It’s moving day for the MNPD Mounted Patrol Unit! The horses will now be housed in a barn at Peeler Park in Madison. As you can see, they’re already enjoying all the space and fresh grass in their new pasture.

Metro Nashville PD

24,406 次观看 • 3 个月前

'LATENT' learns tennis skills for humanoid robots from human motion data. The robot can sustain multi-shot rallies, handle ball speeds of 15+ m/s, and showed a 90.9% success rate for the forehand. No onboard cameras or vision models, relies on external MoCap for high-precision, low-latency ball tracking. Paper:

'LATENT' learns tennis skills for humanoid robots from human motion data. The robot can sustain multi-shot rallies, handle ball speeds of 15+ m/s, and showed a 90.9% success rate for the forehand. No onboard cameras or vision models, relies on external MoCap for high-precision, low-latency ball tracking. Paper:

The Humanoid Hub

63,824 次观看 • 5 个月前

Started using a low-res volumetric cloud map in Wicked Engine for secondary effects. It can be sampled easily by reflections (planar, probe, raytraced) and also helps with disocclusion noise:

Started using a low-res volumetric cloud map in Wicked Engine for secondary effects. It can be sampled easily by reflections (planar, probe, raytraced) and also helps with disocclusion noise:

János Turánszki

17,139 次观看 • 6 个月前

Obstacle courses go far beyond: "just letting kids play" It's unfortunate that we don't have more forward thinking and creative coaches working with young athletes. Often times I see little kids doing little adult training instead! Obstacle courses are a great choice when working with younger pre-adolescent athletes. While many coaches view obstacle courses as “just letting the kids play” or just “running races with things in the way”, the experienced coach knows that any activity can be manipulated to get a desired training effect. Obviously for the youngest of children just “letting them run” an obstacle course at their own pace or racing against a friend is highly engaging and extraordinary for cardiovascular conditioning but we can also use obstacle courses as a task-oriented opportunity to train other elements like mobility, strength, speed, balance and coordination using variety and diversity of movement. Lets take a look at some examples: 1. Mobility obstacle course 2. Crawling/Climbing Obstacle course 3. Motor skills obstacle course 4. Jumping obstacle course A coach can use these exercises as a stand alone activity or combine them in an infinite number of variations to develop a wide range of athletic skills. #LTAD Anyone interested in learning more about youth athletic development DM me for details on courses.

Obstacle courses go far beyond: "just letting kids play" It's unfortunate that we don't have more forward thinking and creative coaches working with young athletes. Often times I see little kids doing little adult training instead! Obstacle courses are a great choice when working with younger pre-adolescent athletes. While many coaches view obstacle courses as “just letting the kids play” or just “running races with things in the way”, the experienced coach knows that any activity can be manipulated to get a desired training effect. Obviously for the youngest of children just “letting them run” an obstacle course at their own pace or racing against a friend is highly engaging and extraordinary for cardiovascular conditioning but we can also use obstacle courses as a task-oriented opportunity to train other elements like mobility, strength, speed, balance and coordination using variety and diversity of movement. Lets take a look at some examples: 1. Mobility obstacle course 2. Crawling/Climbing Obstacle course 3. Motor skills obstacle course 4. Jumping obstacle course A coach can use these exercises as a stand alone activity or combine them in an infinite number of variations to develop a wide range of athletic skills. #LTAD Anyone interested in learning more about youth athletic development DM me for details on courses.

Jeremy Frisch

55,267 次观看 • 2 年前

Mars in motion! ⏩ While on its way to explore a mysterious, metal-rich asteroid, our Psyche spacecraft sped by Mars for a gravity assist to pick up speed and adjust its trajectory. As it approached the planet, it captured this series of high-res snaps.

Mars in motion! ⏩ While on its way to explore a mysterious, metal-rich asteroid, our Psyche spacecraft sped by Mars for a gravity assist to pick up speed and adjust its trajectory. As it approached the planet, it captured this series of high-res snaps.

NASA JPL

70,289 次观看 • 29 天前

Unite will leave no stone unturned in the fight for jobs. We have mounted a significant campaign on this issue and we fully expect the The Labour Party to make a serious commitment to a better future for #UKSteel. Sign here:

Unite will leave no stone unturned in the fight for jobs. We have mounted a significant campaign on this issue and we fully expect the The Labour Party to make a serious commitment to a better future for #UKSteel. Sign here:

Sharon Graham

12,031 次观看 • 2 年前

Robots Using Their Feet to Manipulate Objects ETH Zurich researchers have advanced "pedipulation," where legged robots use their feet for manipulation instead of robotic arms. Their new approach trains robots to navigate and manipulate objects while avoiding static and dynamic obstacles. Tested on the ANYmal quadruped, the method enables precise foot control in complex, unseen environments after minimal training in simulation. Will this research open up new possibilities for efficient, obstacle-aware robot manipulation?

Robots Using Their Feet to Manipulate Objects ETH Zurich researchers have advanced "pedipulation," where legged robots use their feet for manipulation instead of robotic arms. Their new approach trains robots to navigate and manipulate objects while avoiding static and dynamic obstacles. Tested on the ANYmal quadruped, the method enables precise foot control in complex, unseen environments after minimal training in simulation. Will this research open up new possibilities for efficient, obstacle-aware robot manipulation?

Circuit

11,802 次观看 • 1 年前

I've dedicated months to a project that might be entirely a wasted venture. The pool projection thing is a very big ask for each customer. - They need a pool table. - They need a suitable projector mounted above it - They need a suitable camera mounted above it - They need a PC with a GPU - They need to train a model on their balls in their environment Even if it goes no where, it's alive in my studio and it's by far the coolest thing I've ever built, and it's going to turn me into a monster pool player.

I've dedicated months to a project that might be entirely a wasted venture. The pool projection thing is a very big ask for each customer. - They need a pool table. - They need a suitable projector mounted above it - They need a suitable camera mounted above it - They need a PC with a GPU - They need to train a model on their balls in their environment Even if it goes no where, it's alive in my studio and it's by far the coolest thing I've ever built, and it's going to turn me into a monster pool player.

Gary Simon

87,136 次观看 • 3 个月前

NEWS: Hyundai's self-driving unit 42dot has released a new video showing off its end-to-end autonomous driving software on the roads of South Korea, and parking. "It drives using only eight cameras mounted on the vehicle, without relying on an HD maps. It utilizes a 400 TOPS-class NPU to implement the NOA (Navigation on Autopilot) function on highways and in urban areas."

NEWS: Hyundai's self-driving unit 42dot has released a new video showing off its end-to-end autonomous driving software on the roads of South Korea, and parking. "It drives using only eight cameras mounted on the vehicle, without relying on an HD maps. It utilizes a 400 TOPS-class NPU to implement the NOA (Navigation on Autopilot) function on highways and in urban areas."

Sawyer Merritt

506,271 次观看 • 8 个月前

Here's Elon Musk in 2020 explaining why cameras with AI vision is all you need to achieve full autonomy. “The entire road network is designed for passive optical, essentially vision. In order to make a car drive properly, you have to solve vision. At the point at which you solve vision, you really don't need any other instruments. A careful human driver can drive with an extremely good track record. Unlike a human, the computer does not get tired, it has 360-degree surround cameras, it's got three cameras pointing forward. It's like being able to see with eyes in the back of your head. Vision is the way to go. There's some value to active optical for a wavelength that's occlusion-penetrating, so it can see through fog, or rain, or dust. But it has to be high resolution, such that you can rely on it, for example, for radar at roughly a four-millimeter wavelength. This is good for occlusion penetration. But it needs to have enough resolution to know that you're braking for a real object and not just a bridge or a manhole cover or something like that.” Axel Springer Pioneers Talk, December 1, 2020

Here's Elon Musk in 2020 explaining why cameras with AI vision is all you need to achieve full autonomy. “The entire road network is designed for passive optical, essentially vision. In order to make a car drive properly, you have to solve vision. At the point at which you solve vision, you really don't need any other instruments. A careful human driver can drive with an extremely good track record. Unlike a human, the computer does not get tired, it has 360-degree surround cameras, it's got three cameras pointing forward. It's like being able to see with eyes in the back of your head. Vision is the way to go. There's some value to active optical for a wavelength that's occlusion-penetrating, so it can see through fog, or rain, or dust. But it has to be high resolution, such that you can rely on it, for example, for radar at roughly a four-millimeter wavelength. This is good for occlusion penetration. But it needs to have enough resolution to know that you're braking for a real object and not just a bridge or a manhole cover or something like that.” Axel Springer Pioneers Talk, December 1, 2020

ELON CLIPS

1,513,211 次观看 • 1 年前

New preprint - basically we noticed that by applying ab initio (SDG) at high res in CS, using a very small step size, we could crack some small proteins we hadn't been able to solve before. Read the preprint or head over to bluesky for a longer thread!

New preprint - basically we noticed that by applying ab initio (SDG) at high res in CS, using a very small step size, we could crack some small proteins we hadn't been able to solve before. Read the preprint or head over to bluesky for a longer thread!

Oli B. Clarke

23,781 次观看 • 11 个月前